Vehicle collision detection method, system, equipment and medium

Abstract

The invention discloses a vehicle collision detection method, system and device and a medium. The method comprises the steps that S1, high-frequency data of a vehicle acceleration sensor are collected; s2, calculating an acceleration range sum at the current moment, and if the range sum is greater than a preset threshold value, intercepting high-frequency packet data in front and back set time periods with the current moment as the center; s3, the TICC model is improved and trained, acceleration time history data intercepted in a set time period is segmented and classified, and sub-intervals of a collision duration period are obtained; s4, extracting artificial feature vectors of the collision subintervals, and training a fuzzy CBLOF model; and S5, carrying out abnormal degree scoring on the feature vectors, and if the score is greater than a set threshold value, giving a collision alarm. According to the method, better segmentation and clustering are carried out on accelerator sensor data based on the improved TICC algorithm, feature scoring is carried out by adding the CBLOF model of the comprehensive fuzzy evaluation matrix and fusing expert experience, and the precision of vehicle collision detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle collision detection, and in particular to a vehicle collision detection method, system, equipment and medium. Background Art

[0002] With the rapid development of the automobile industry, the number and types of cars are increasing, and the incidence of traffic accidents is also rising. In order to improve the safety performance of automobiles, collision detection technology has been widely studied, applied and developed. The vehicle collision detection system can respond quickly at the moment of an accident, effectively reducing casualties and property losses.

[0003] At present, vehicle collision detection technology is mainly a detection system that combines sensors and artificial intelligence algorithms. Common sensors include accelerometers, pressure sensors, ultrasonic sensors, radars, and cameras. When a vehicle collides, these sensors automatically detect the collision event by collecting real-time data from the vehicle and cooperating with the algorithm system, and send the collision information to the emergency rescue center through the vehicle communication system (such as GPS, cellular network). The collision information usually includes the vehicle location, the severity of the collision, the direction of the collision, and the occupant information on the vehicle. After receiving the information, the rescue center will immediately contact the nearest emergency department and dispatch rescue personnel to the scene of the accident. Users can also track the rescue progress in real time through mobile phone applications or vehicle-mounted systems, and provide feedback after the rescue is completed. This feedback information is also of great reference value for OEMs to improve collision detection and rescue services.

[0004] The collision detection method based on accelerometer has its unique advantages in the field of vehicle collision detection. Compared with other methods, its advantages are mainly reflected in the following aspects: when a collision occurs, the vehicle will produce an instantaneous and drastic acceleration change. The accelerometer can capture this change within milliseconds, thereby quickly triggering the collision detection system; and compared with other high-precision sensors (such as radar and lidar), the manufacturing cost of the accelerometer is relatively low and can be mass-produced. This makes it easier to promote and apply the collision detection system based on accelerometers in various types of vehicles, including economical models; on the other hand, the accelerometer is usually small in size and easy to install in various locations of the vehicle. Its installation and maintenance are relatively simple and do not require a complicated calibration process. Compared with the multi-sensor system, the system based on accelerometer reduces the workload of maintenance and calibration, and reduces the cost of use.

[0005] Current accelerometers usually have multi-axis detection capabilities and can monitor the acceleration changes of vehicles in multiple directions. This enables the system to detect collisions in different directions, including frontal, side and rear collisions, and provide comprehensive collision protection. Although current accelerometers can collect multivariate data, the current vehicle collision detection methods cannot process multivariate time series data well. Affected by the data of non-collision time series, the accuracy of vehicle collision detection is not high. Since the time series clustering problem of multivariate time series is more difficult than the standard time series segmentation, it is necessary to implement the segmentation function and identify the recurring patterns; this problem is also more difficult than subsequence clustering because it requires clustering adjacent timestamps into one category as much as possible. In addition, even if the multivariate time series can be segmented and clustered at the same time, it is a very challenging problem to generate a highly interpretable representation of the clustering results. Therefore, an unsupervised segmentation and clustering algorithm is needed to solve this problem and accurately capture the characteristics of the accident fragments.

[0006] Toeplitz Inverse Covariance-Based Clustering (TICC) is a clustering algorithm for high-dimensional time series data. The algorithm combines the characteristics of sparse inverse covariance estimation and Toeplitz structure, and can effectively identify and cluster different patterns or states in time series. It is widely used in financial data analysis, biomedical signal processing and other fields, but has not been effectively applied in vehicle collision detection. Summary of the invention

[0007] To solve the above problems, the present invention provides a vehicle collision detection method, system, device and medium. The method performs better segmentation and clustering of accelerator sensor data based on an improved TICC algorithm. At the same time, the CBLOF model with a comprehensive fuzzy evaluation matrix is ​​added to integrate expert experience, and a greater weight is given to features that can better represent the collision, so that anomaly scores can be output more accurately at the feature level, thereby improving the accuracy of vehicle detection.

[0008] The present invention provides a vehicle collision detection method, and the specific technical solution is as follows:

[0009] S1: Collect high-frequency data from the vehicle acceleration sensor, including the acceleration components of three axes;

[0010] S2: Calculate the acceleration range sum at the current moment and determine whether it is greater than a preset threshold. If so, intercept the high-frequency packet data of the set time period before and after the current moment;

[0011] S3: Improve and train the TICC model, segment and classify the acceleration time history data captured in a set time period based on the trained TICC model, and obtain the sub-intervals of the collision duration period;

[0012] S4: extract artificial feature vectors of collision subintervals and train fuzzy CBLOF model;

[0013] S5: The feature vector is scored for abnormality based on the trained fuzzy CBLOF model. If the abnormality score is greater than the set threshold, a collision alarm is issued.

[0014] Furthermore, in step S2, the acceleration range sum at the current moment is calculated as follows:

[0015] Calculate the range of acceleration of the x, y, and z axes per second, and sum the ranges of the three axes to obtain the sum of the acceleration ranges at the current moment.

[0016] Furthermore, in step S3, the improvement of the TICC model is as follows:

[0017] Define the symmetric Toeplitz matrix Θ i Describing the correlation between sensors, the symmetric Toeplitz matrix Θ i , which is expressed as follows:

[0018]

[0019] Among them, A (0) …A (4) ∈R 3×3 , Indicates the dependency between the i-axis and j-axis of the acceleration sensor at the current moment. Indicates the dependency between the i-axis at the current moment and the j-axis at the next z moments.

[0020] Furthermore, the TICC model is improved and the training process is as follows:

[0021] S301: Acquire the axis with the largest absolute value of the acceleration sensor in a stationary state as the gravity axis;

[0022] S302: Acquire high-frequency data of three-axis acceleration with a duration equal to the duration of the intercepted time period;

[0023] S303: Setting a sliding time window to obtain a data matrix of an input model;

[0024] S304: Initialize the inverse covariance parameters corresponding to each category, and fix these parameters, update the segmentation set P, and after updating the segmentation set P, fix P, update the inverse covariance, and iterate the update of the segmentation set P and the inverse covariance until the segmentation clustering result converges or reaches the set number of iterations, and output the segmentation set parameters P and the inverse covariance matrix parameters.

[0025] Furthermore, in step S4, the fuzzy CBLOF model is independently trained according to a single feature based on the artificial feature vector.

[0026] Furthermore, in step S5, the feature vector is scored for abnormality, and the specific process is as follows:

[0027] S501: Score the importance of each feature, classify the degree of abnormality, and construct a fuzzy relationship matrix;

[0028] S502: Use the trained CBLOF model to score the abnormality of each feature of the new data, and normalize the abnormality score values ​​of all feature dimensions to obtain a fuzzy set;

[0029] S503: performing matrix multiplication operation on the fuzzy set and the fuzzy relationship matrix to obtain a fuzzy comprehensive evaluation result vector of the object to be evaluated;

[0030] S504: assigning different weights to abnormality levels according to the abnormality levels, and calculating the final fuzzy comprehensive evaluation score.

[0031] Furthermore, the final fuzzy comprehensive evaluation score is calculated as follows:

[0032]

[0033] Among them, b j Represents the jth element of the fuzzy comprehensive evaluation result vector of the object to be evaluated.

[0034] The present invention also provides a vehicle collision detection system. Based on the above method, the system includes an acceleration sensor, a server, a model detection and analysis module, and a display output unit;

[0035] The acceleration sensor is used to collect high-frequency acceleration data of the three axes of the vehicle and transmit it to the server;

[0036] The server includes a vehicle-mounted server or a cloud server, receives the three-axis acceleration data, and calculates the acceleration range sum at the current moment, and determines whether it is greater than a preset threshold value. If it is greater, the high-frequency packet data of the set time period before and after the current moment is intercepted from the received data;

[0037] The model detection and analysis module receives the intercepted high-frequency packet data, and scores the abnormality based on the above method. If the abnormality score is greater than a set threshold, an alarm signal is sent to the display output unit;

[0038] The display output unit receives the alarm signal and issues collision alarm information.

[0039] The present invention also provides a vehicle collision detection device, which includes: a memory, a processor, and a vehicle collision detection program stored in the memory and executable on the processor. When the vehicle collision detection program is executed by the processor, the steps of the above-mentioned vehicle collision detection method are implemented.

[0040] The present invention also provides a computer storage medium, on which a vehicle collision detection program is stored. When the vehicle collision detection program is executed by a processor, the steps of the vehicle collision detection method described above are implemented.

[0041] The beneficial effects of the present invention are as follows:

[0042] 1. The method of the present invention can realize vehicle collision detection only through three-axis acceleration, which greatly saves hardware costs compared with the existing collision detection method that needs to collect multiple sensor data. At the same time, compared with the existing image-based collision detection method that can only detect the direction that the lens can detect, the vehicle collision detection method of the present invention does not have an obvious blind spot in the field of vision and can realize all-round collision detection.

[0043] 2. The present invention adopts an improved TICC model, which can be trained on unlabeled data and optimizes the feature selection of the car collision scene. Therefore, it can accurately capture the acceleration time history of the accident fragment, and can be more accurate in the calculation of accident characteristics without being affected by the data of non-collision time history.

[0044] 3. The present invention adopts a fuzzy CBLOF algorithm to score the degree of abnormality of the evaluation object. By adding a comprehensive fuzzy evaluation matrix, the CBLOF model can well integrate expert experience, score each feature according to different levels, give greater weight to features that can better represent collisions, and output abnormality scores more accurately at the feature level. The final comprehensive score can well characterize whether the vehicle has collided, thereby improving the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the overall process of the method of the present invention.

[0046] Figure 2 It is a schematic diagram of the improved TICC model training process of the present invention.

[0047] Figure 3 It is a schematic diagram of the process of scoring the abnormality degree of a feature vector according to the present invention.

[0048] Figure 4 It is a schematic diagram of the overall system architecture of the present invention. DETAILED DESCRIPTION

[0049] The following description clearly and completely describes the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In the description of the embodiments of the present invention, it should be noted that the indicated orientation or positional relationship is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the invention product is usually placed when in use, or the orientation or positional relationship commonly understood by those skilled in the art, or the orientation or positional relationship in which the invention product is usually placed when in use, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first" and "second" are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0051] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] Example 1

[0053] Embodiment 1 of the present invention discloses a vehicle collision detection method, such as Figure 1 As shown, the specific process is as follows:

[0054] S1: Collect high-frequency data from the vehicle acceleration sensor, including the acceleration components of three axes;

[0055] Specifically, set the acquisition frequency to fhz to obtain the acceleration components of the x, y, and z axes;

[0056] S2: Calculate the acceleration range sum at the current moment and determine whether it is greater than a preset threshold. If so, intercept the high-frequency packet data of the set time period before and after the current moment;

[0057] As a preferred embodiment, the acceleration range sum at the current moment is calculated as follows:

[0058] Calculate the range (maximum value - minimum value) of the acceleration of the x, y, and z axes per second, and add the ranges of the three axes to obtain the acceleration range sum S at that moment;

[0059] The preset threshold is denoted as a, and the set time period is denoted as n;

[0060] If the sum of the three-axis ranges S is greater than the threshold a, the high-frequency packet data of n seconds before and after the current moment is intercepted.

[0061] S3: Improve and train the TICC model, segment and classify the n-second acceleration time history data based on the trained TICC model, and obtain the sub-intervals of the collision duration;

[0062] Each category obtained by TICC clustering represents a specific state in the three-axis time series of the acceleration sensor. When the three-axis time series is in this specific state, a certain time-invariant structure will be retained. This structure will exist throughout the time window covered by the entire category, regardless of the position where the time window begins.

[0063] Based on this, the improvements of the TICC model are as follows:

[0064] Define the symmetric Toeplitz matrix Θ i Describing the correlation between sensors, the symmetric Toeplitz matrix Θ i , which is expressed as follows:

[0065]

[0066] Among them, A (0) …A (4) ∈R 3×3 , Indicates the dependency between the i-axis and j-axis of the acceleration sensor at the current moment. Indicates the dependency between the i-axis at the current moment and the j-axis at the next z moments.

[0067] Θ i A Markov random field of clustering categories is defined. The goal of the algorithm is to solve K inverse covariance matrices Θ = {Θ1, ..., Θ K} (K is the number of clusters), and the segmentation set P = {P1, ..., P K},in, T is the total duration.

[0068] As a preferred embodiment, Figure 2 As shown, the model training process is as follows:

[0069] S301: Acquire the axis with the largest absolute value of the acceleration sensor in a stationary state as the gravity axis;

[0070] S302: Acquire sufficient high-frequency data of three-axis acceleration with a duration of n seconds, each high-frequency data is V = [v0, v1, ..., v fn ]∈R 3×fn , where v i =[x i ,y i ,z i ] T ;

[0071] S303: Set a sliding time window with a length of w and a stride of 1. Each window vector V i =[V i-w+1 ,…,V i ] as the minimum data granularity of the clustering algorithm, obtain the data matrix V of the input model w =[V w ,…,V fn ]∈R 3w×(fn-w+1) ;

[0072] S304: Initialize the inverse covariance parameters corresponding to each category, fix these parameters, and update the segmentation set P;

[0073] Since the real collision is more accurate on the forward axis and the steering axis, and the fluctuation of the gravity axis is mostly caused by normal driving, only the characteristic components of the two dimensions of the forward axis and the steering axis are input. At this stage, the sub-problem is transformed into the following objective function:

[0074]

[0075] Among them, the input characteristic parameters V t ,Θ i A variant with only non-gravity axis components, To ensure that the vectors at adjacent moments are clustered into one category as much as possible and to maintain the time consistency parameters, is the likelihood function value.

[0076] After updating the segmentation set P, fix P and update the inverse covariance;

[0077] Updating the inverse covariance is to extract the mode of the overall acceleration in three axes, so the complete V t Input, in the objective function Since P is fixed and becomes a variable C, it can be omitted. In this way, the objective function of the sub-problem at this stage will be defined as follows:

[0078]

[0079] Among them, |P i | is the number of categories belonging to i, S i is the covariance of these points.

[0080] Dynamic programming and ADMM algorithms are used to solve the above two problems in turn, and the segmentation set P and inverse covariance are updated cyclically and iteratively until the segmentation clustering results converge or the set number of iterations is reached, and the segmentation set parameters P and inverse covariance matrix parameters are output.

[0081] S4: extract artificial feature vectors of collision subintervals and train fuzzy CBLOF model;

[0082] The artificial feature vector includes a range feature, a variance feature, an information entropy feature and a slope feature;

[0083] Specifically, based on the artificial feature vector, the fuzzy CBLOF model is trained independently according to a single feature.

[0084] By improving the TICC model clustering, the most abnormal collision sub-interval is obtained, and the collision characteristics in this interval are calculated, so that after obtaining the three-axis acceleration intervals at the beginning and end of the collision, the collision characteristics can be calculated more accurately without being affected by the data in the non-collision area. The range feature is the maximum acceleration minus the minimum acceleration, the variance feature and information entropy are the variance and information entropy of the acceleration in the collision interval, and the slope feature is the absolute value of the maximum first-order difference of the acceleration in the collision interval.

[0085] S5: The feature vector is scored for abnormality based on the trained fuzzy CBLOF model. If the abnormality score is greater than the set threshold, a collision alarm is issued.

[0086] Score the abnormality of the feature vector, such as Figure 3 As shown, the specific process is as follows:

[0087] S501: Based on expert experience, score the importance of each feature, and classify the abnormality into n levels, and construct a fuzzy relationship matrix R∈(m,n), where m is the number of features and n is the number of abnormality levels;

[0088] S502: Use the trained CBLOF model to score the abnormality of each feature of the new data, and normalize the abnormality scores of all feature dimensions to obtain the fuzzy set T i , as follows:

[0089]

[0090] T i =(s1',s'2,...,s' m)

[0091] S503: The fuzzy set T i Perform matrix multiplication operation with the fuzzy relationship matrix R to obtain the fuzzy comprehensive evaluation result vector B of the object to be evaluated, which is expressed as follows:

[0092]

[0093] The traditional CBLOF model has no way to assign different weights to different features and treat all features equally. However, after adding the comprehensive fuzzy evaluation matrix, the CBLOF model can well integrate expert experience, give greater weights to features that are more representative of collisions, output anomaly scores more accurately at the feature level, and improve detection accuracy.

[0094] S504: assign different weights to different levels of abnormality according to the level of abnormality, and calculate the final fuzzy comprehensive evaluation score, which is calculated as follows:

[0095]

[0096] Among them, b j Represents the jth element of the fuzzy comprehensive evaluation result vector of the object to be evaluated.

[0097] Example 2

[0098] Embodiment 1 of the present invention discloses a vehicle collision detection system, based on the vehicle collision detection method described in embodiment 1 above, such as Figure 4 As shown, the system includes an acceleration sensor, a server, a model detection and analysis module, and a display output unit;

[0099] The acceleration sensor is used to collect high-frequency acceleration data of the three axes of the vehicle and transmit it to the server;

[0100] The server includes a vehicle-mounted server or a cloud server, receives the three-axis acceleration data, and calculates the acceleration range sum at the current moment, and determines whether it is greater than a preset threshold value. If it is greater, the high-frequency packet data of the set time period before and after the current moment is intercepted from the received data;

[0101] The model detection and analysis module receives the intercepted high-frequency packet data, and performs an abnormality score based on the method described in the above embodiment 1, and sends an alarm signal to the display output unit if the abnormality score is greater than a set threshold;

[0102] The display output unit receives the alarm signal and issues collision alarm information.

[0103] Example 3

[0104] Embodiment 3 of the present invention discloses a vehicle collision detection device, which may be a user equipment (UE), such as a mobile phone, a smart phone, a laptop computer, a digital broadcast receiver, a personal digital assistant (PDA), a tablet computer (PAD), a handheld device, a vehicle-mounted device, a wearable device, a computing device or other processing device connected to a wireless modem, a mobile station (MS), etc., for executing a vehicle collision detection method. The device may be called a user terminal, a portable terminal, a desktop terminal, etc.

[0105] Typically, the device includes: at least one processor, a memory, and a vehicle collision detection program stored in the memory and executable on the processor, wherein the vehicle collision detection program is configured to implement the steps of the vehicle collision detection method as described in Example 1.

[0106] The processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. The processor may also include an AI (Artificial Intelligence) processor, which is used to process calculation operations related to the vehicle collision detection program, so that the vehicle collision detection method can be trained and learned autonomously to improve efficiency and accuracy.

[0107] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory is used to store at least one instruction, which is used to be executed by the processor to implement the vehicle collision detection method described in the above embodiment 1.

[0108] In some embodiments, the terminal may also optionally include: a communication interface and at least one peripheral device. The processor, the memory and the communication interface may be connected via a bus or a signal line. Each peripheral device may be connected to the communication interface via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit, a display screen and a power supply.

[0109] The communication interface can be used to connect at least one peripheral device related to I / O (Input / Output) to the processor and the memory. The communication interface is used to receive the movement trajectories and other data of multiple mobile terminals uploaded by users through the peripheral device. In some embodiments, the processor, memory and communication interface are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor, memory and communication interface can be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0110] The radio frequency circuit is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit communicates with the communication network and other communication devices through electromagnetic signals, so that the movement trajectory and other data of multiple mobile terminals can be obtained. The radio frequency circuit converts the electrical signal into an electromagnetic signal for transmission, or converts the received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like. The radio frequency circuit can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to: a metropolitan area network, various generations of mobile communication networks (2G, 3G, 4G and 5G), a wireless local area network and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit may also include circuits related to NFC (Near Field Communication), which are not specifically limited here.

[0111] The display screen is used to display the UI (User Interface). The UI may include graphics, text, icons, videos and any combination thereof. When the display screen is a touch display screen, the display screen also has the ability to collect touch signals on the surface or above the surface of the display screen. The touch signal can be input to the processor as a control signal for processing. At this time, the display screen can also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards. In some embodiments, the display screen can be one, the front panel of the electronic device; in other embodiments, the display screen can be at least two, respectively arranged on different surfaces of the electronic device or in a folding design; in some further embodiments, the display screen can be a flexible display screen, arranged on a curved surface or a folding surface of the electronic device. Even, the display screen can also be set to a non-rectangular irregular shape, that is, a special-shaped screen. The display screen can be made of materials such as LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode) and the like.

[0112] The power supply is used to power various components in the electronic device. The power supply can be AC, DC, a disposable battery or a rechargeable battery. When the power supply includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0113] Example 4

[0114] Embodiment 4 of the present invention discloses a computer storage medium, which is a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the vehicle collision detection method of the above method embodiment are implemented.

[0115] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0116] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A vehicle collision detection method, characterized in that: include: S1: Collect high-frequency data from the vehicle acceleration sensor, including the acceleration components of three axes; S2: Calculate the acceleration range sum at the current moment and determine whether it is greater than a preset threshold. If so, intercept the high-frequency packet data of the set time period before and after the current moment; S3: Improve and train the TICC model, segment and classify the acceleration time history data captured in a set time period based on the trained TICC model, and obtain the sub-intervals of the collision duration period; S4: extract artificial feature vectors of collision subintervals and train fuzzy CBLOF model; S5: The feature vector is scored for abnormality based on the trained fuzzy CBLOF model. If the abnormality score is greater than the set threshold, a collision alarm is issued.

2. The vehicle collision detection method according to claim 1, characterized in that: In step S2, the acceleration range sum at the current moment is calculated as follows: Calculate the range of acceleration of the x, y, and z axes per second, and sum the ranges of the three axes to obtain the sum of the acceleration ranges at the current moment.

3. The vehicle collision detection method according to claim 1, characterized in that: In step S3, the improvement of the TICC model is as follows: Define the symmetric Toeplitz matrix Θ i Describing the correlation between sensors, the symmetric Toeplitz matrix Θ i , which is expressed as follows: Among them, A (0) …A (4) ∈R 3×3 , Indicates the dependency between the i-axis and j-axis of the acceleration sensor at the current moment. Indicates the dependency between the i-axis at the current moment and the j-axis at the next z moments.

4. The vehicle collision detection method according to claim 3, characterized in that: Improve the TICC model, the training process is as follows: S301: Acquire the axis with the largest absolute value of the acceleration sensor in a stationary state as the gravity axis; S302: Acquire high-frequency data of three-axis acceleration with a duration equal to the duration of the intercepted time period; S303: Setting a sliding time window to obtain a data matrix of an input model; S304: Initialize the inverse covariance parameters corresponding to each category, and fix these parameters, update the segmentation set P, and after updating the segmentation set P, fix P, update the inverse covariance, and iterate the update of the segmentation set P and the inverse covariance until the segmentation clustering result converges or reaches the set number of iterations, and output the segmentation set parameters P and the inverse covariance matrix parameters.

5. The vehicle collision detection method according to claim 1, characterized in that: In step S4, based on the artificial feature vector, the fuzzy CBLOF model is independently trained according to a single feature.

6. The vehicle collision detection method according to claim 5, characterized in that: In step S5, the feature vector is scored for abnormality, and the specific process is as follows: S501: Score the importance of each feature, classify the degree of abnormality, and construct a fuzzy relationship matrix; S502: Use the trained CBLOF model to score the abnormality of each feature of the new data, and normalize the abnormality score values ​​of all feature dimensions to obtain a fuzzy set; S503: performing matrix multiplication operation on the fuzzy set and the fuzzy relationship matrix to obtain a fuzzy comprehensive evaluation result vector of the object to be evaluated; S504: assigning different weights to abnormality levels according to the abnormality levels, and calculating the final fuzzy comprehensive evaluation score.

7. The vehicle collision detection method according to claim 6, characterized in that: The final fuzzy comprehensive evaluation score is calculated as follows: Among them, b j Represents the jth element of the fuzzy comprehensive evaluation result vector of the object to be evaluated.

8. A vehicle collision detection system, characterized in that: Based on the vehicle collision detection method described in any one of claims 1 to 7, the system includes an acceleration sensor, a server, a model detection and analysis module, and a display output unit; The acceleration sensor is used to collect high-frequency acceleration data of the three axes of the vehicle and transmit it to the server; The server includes a vehicle-mounted server or a cloud server, receives the three-axis acceleration data, and calculates the acceleration range sum at the current moment, and determines whether it is greater than a preset threshold value. If it is greater, the high-frequency packet data of the set time period before and after the current moment is intercepted from the received data; The model detection and analysis module receives the intercepted high-frequency packet data, and performs an abnormality score based on the vehicle collision detection method according to any one of claims 1 to 7, and sends an alarm signal to the display output unit if the abnormality score is greater than a set threshold; The display output unit receives the alarm signal and issues collision alarm information.

9. A vehicle collision detection device, characterized in that: The vehicle collision detection device includes: a memory, a processor, and a vehicle collision detection program stored in the memory and executable on the processor. When the vehicle collision detection program is executed by the processor, the steps of the vehicle collision detection method described in any one of claims 1 to 7 are implemented.

10. A computer storage medium, characterized in that: The storage medium stores a vehicle collision detection program, and when the vehicle collision detection program is executed by the processor, the steps of the vehicle collision detection method described in any one of claims 1-7 are implemented.

Images